One of my primary projects is developing and maintaining a reporting datamart for the accounting department. When a report is run, a dataset containing key and fact columns are queried from the corporate ODS (operational data store) into what we call interface tables and retained there for archival purposes. Each reporting dataset is keyed on a report header table with columns that include the runtime parameters, time the query started, when it completed, and text of the SQL SELECT statement. When called upon, we can go back and retreive the exact dataset for any report produced in the past as well as answer questions about it's runtime performance or changes in SQL coding over time. That's a technique that I developed early in my career and it has proven it's worth on many projects. The key point here is that we're not tracking every change that occurs in the ODS (that's more or less maintained by the ODS team), the accounting datamart only archives data aggregated for the department's purpose, so it's kept relatively small and manageble.